TECHNOLOGY MANAGEMENT 3be- TECHNOLOGICAL .ppt

1、TECHNOLOGY MANAGEMENT 3be: TECHNOLOGICAL FORECASTING Essentials,CP Beukman Engineering Management Programme University of Canterbury Issue 1.0 April 2000,Think of Fairy Tales,Jack and the beanstalkSpace ShuttleSeven League BootsBoeing 747Magic MirrorTelevision / Internet,Ask what people want,“We wan

2、t to be able to”,Objectives,Introduce TF methods Application to R&D Measures of technology,TIME,PARAMETER OF PERFORMANCE,LIMIT OF PERFORMANCE DUE TO A NATURAL CONSTRAINT,Embryonic,TECHNOLOGICAL MATURITY,Growth,Mature,Aging,Methods of forecasting,Extrapolation Leading indicators Causal models Probabi

3、listic methods Expert opinion,Extrapolation Methods,Assume the process which produced the historical outcomes will continue to produce future outcomes. Growth curves apply to single technical approaches involve an upper limit Trends apply to overall technologies,Growth Curves,Types growth aids furth

4、er growth upper limit is the only influenceApplication performance extent of adoption,Growth Curve Equations Pearl Curve,Equation:-bt y = L /(1 + ae )y = value L = upper limit a,b = numerical fitted coefficients extract a,b from historic data, not best fit,Growth Curve Equations Gompertz curve,Equat

5、ion:-kt-bey = LeIf y is very small, the curve becomes almost exponential,Growth Curves of Performance,Apply to a single technical approachUpper limit is set by physics & chemistry of technical approach,Year,1,1880,1900,1920,1940,1960,10,100,Efficiency of Incandescent Lamps Gompertz Curve Fitted,Lume

6、ns/Watt,Semi Log plot No accelerating factor (Current performance makes development easier),Year,10,1900,1910,1920,1930,1940,100,1000,Aircraft Speed 1906 - 1925,Speed m/h,Official speed records All a/c wood & fabric Upper limit set by structural strength U/L = 350 mph,Base 10 Pearl Curve,A - BtY = L

7、 / ( 1 + 10 )Fisher-Pry TransformationA - BtY / ( L - Y ) = 10Grows by a factor of 10 in 1/B years,Year,0.01,1904,1908,1912,1916,1920,0.1,1,Aircraft Speed 1906 - 1925,Fisher-Pry Ratio,1924,10,1E+2,The trend assumes thatthe past will be replicatedin the future,Extent of Adoption,Forecast fraction of

8、applications using the technology Typically follows Pearl Curve Early adoptions and further adoption Last holdouts represent hard-to-fill applications Commonly displayed as Fisher-Pry curve,Year,0,1970,1980,1990,50,100,Microwave Ovens in US Households,Percent Households with microwave ovens,2000,10,

9、20,30,40,60,70,80,90,Year,0.01,1970,1975,1980,1985,1990,0.1,1,Microwave Ovens in US Households,Log Number of Households,1995,10,100,Year,1999,2000,2001,2002,Households with structured wiring in the USA,Thousands of Households,2003,0,500,100,200,300,400,600,700,800,900,Year,1E+3,1999,2000,2001,2002,1

10、E+4,1E+5,Households with structured wiring in the USA,Number of households -log scale,2003,1E+6,1E+7,Year,0.00,1985,1990,1995,2000,0.01,0.10,Adoption of Anti-Lock Brakes USA Vehicles,Cars with ABS / Cars without ABS,1.00,10.00,100.00,1000.00,Year,0.01,0.10,Fisher-Pry Ratio,1.00,10.00,1975,1980,1985,

11、1990,1995,Cassette Tapes Compact Disks,Music Recording,Choice of Growth Curves,Forecasting Performance Unexploited potential in early improvements: Pearl Curve No unexploited potential: Gompertz Curve Forecasting Adoption Imitation aids adoption: Pearl Curve Existing adoption aids further adoption:

12、Pearl Curve Otherwise: Gompertz Curve,Long Term Forecasts,Growth curves cannot project beyond the upper limit of the technical approach There is often a need to prepare a forecast which goes beyond the current technical approach Trend projection is how this is done,How Will Trend be Extended?,Some n

13、ew technical approach will be needed New approach will have a higher upper limit than the one currently used It is not the job of the forecaster to identify what this new approach will (actually) be Sufficient to warn that it will come,Trends,1. No imminent progress 2. Growth proportional to progres

14、s already made (exponential growth)ktY = y0 eln Y = ln y0 + ktlog Y = log y0 + kt log e,10,100,Speed mph,1000,10,000,1900,1920,1940,1960,1980,Wood & fabric,All metal,Subsonic jets,Aircraft speed, growth curves and trend,Supersonic jets,100,Maximum Speed mph,1000,10,000,1920,1940,1960,1980,Transport

15、aircraft speed trend,Combat Aircraft Speed Trend,Mach 2.2,Speed Trends of Combat vs Transport Aircraft, Showing Lead Trend Effect,Mach 2.7,2000,Source: Ralph C Lenz, Forecasts of Exploding Technologies by Trend Extrapolation,Combat Aircraft,Transport Aircraft,1,10,Nanoseconds,100,1000,1970,1980,1990

16、,2000,Clock Time of Microprocessors and Vector Processors,MIPS R2000,Cray 1S,Cray X-MP,Cray Y-MP,Cray C90,MIPS R3000,HP 7000,R4000,R4000,DEC Alpha,0.1,1.0,Minimum Feature Size Microns,10.0,100.0,1970,1980,1990,2000,Minimum Feature Size Dynamic Random Access Memories,Moores Law,1e-4,1e-1,Seconds,1e+2

17、,1820,1880,1900,1980,Exposure time for Photographic Film f16 in Bright Sunlight,1e-3,1e-2,1e+0,1e+1,1e+3,1e+4,1840,1920,1940,1960,1860,1830 exposure = 30 minutes Now around 1/10,000 sec No comment on graininess Film not optimised for illumination source,10,Bushels/Acre,1800,1920,1960,Yield of Corn o

18、n US Farms Effect of an outside influence,100,1840,2000,1880,Trend break in 1930s Driver = politics Before 1930 - increase = more land or more horses 1930 higher prices & limitation on land subdivision + chemicals, fertilisers, weedkillers,1929 drought,Current Dollars,1750,1900,1950,USA Management C

19、apability Dollar Magnitude of Engineering Projects,1800,2000,1850,1e+4,1e+7,1e+10,1e+5,1e+6,1e+8,1e+9,1e+11,Canals,Roads/Rail,Electrical,Bridges,Space/Nuclear,Petroleum,Miscellaneous,Trends: Summary,Used to project beyond the limits of current technical approach Based on series of successive technic

20、al approaches Growth in performance is usually exponential Qualitative trends can be developed if necessary,Leading Indicators,Precursors of coming events May be events of different nature from event to be forecast Succession of leading indicators gives additional strength to the forecast,Stages of

21、Innovation,Scientific findings Laboratory feasibility Operating prototype Commercial introduction/operational use Widespread adoption Diffusion to other areas Social & economic impact,Automotive Innovations,Normal sequence of innovation: demonstration car prestige car mass market car,Lag Times for A

22、utomotive Innovations,0,2,4,6,8,10,12,14,16,Turbocharger,Plastic structural parts,Prestige - mass,Demo - Prestige,Plastic Body Shell,Fuel Injection,Electronic Ignition,Electronic Engine Ctrl,Automobile technology,Turbo - migrated from dieselsFuel injectors ability to manufacture lower cost injectors

23、 (supporting technology in the manufacturing process),Military Technology,In the past civilian technology has led military rifled guns - sporting before military machine guns - resisted by militarySince 1930s military has led civilian, especially in aerospaceNow changing again,100,Maximum Speed mph,

24、1000,10,000,1920,1940,1960,1980,Transport aircraft speed trend,Combat Aircraft Speed Trend,Mach 2.2,Speed Trends of Combat vs Transport Aircraft, Showing Lead Trend Effect,Mach 2.7,2000,Source: Ralph C Lenz, Forecasts of Exploding Technologies by Trend Extrapolation,Fighter Aircraft,Transport Aircra

25、ft,Bomber Aircraft,Transistor,1926 Lilienfield applies for patent on solid state amplifier (granted 1930) thus FET described 1936 Heil granted patent on solid state amplifier - - - - - - - - - - - - - - - - (no pure materials) 1947 Point contact transistor 1951 Junction transistor 1952 Field Effect

26、Transistor,Penicillin,Folk medicine: mint leaves and sugar spread on open wound 1897 Duchesne thesis on use of mold extract to cure disease 1928 Fleming observes anti-bacterial effect of mold - - - - - - - - - - - - - 1939 Penicillin isolated by Chaim & Florey,Precursors: Summary,Lead-lag relation w

27、ill permit forecasting Precursors of innovation can be sought systematically Multiple indicators increase confidence in forecast Thresholds must be set to balance risks of false warning, warning too late,Probabilistic Methods,Other methods give single point forecast Probabilistic methods give range

28、of outcomes, probability distribution over the outcomes Not yet fully developed or widely used,Types of Forecasts,Stochastic Projection Probabilistic Lags Bayesian Updating with Multiple Precursors,Stochastic Projection,Use historic data to generate probability distributions Obtain probability distr

29、ibutions: time intervals performance increases Simulate growth using time steps and performance increases drawn from the historical probability distributions,Time Between Aircraft Speed Records,Probability,Probability of Speed Increase,Longer times between attempts mean greater technology advance Lo

30、nger times between attempts mean greater speed increase “Correlated” portion of time interval and speed increase must be calculated as a deterministic value Random portion of speed increase added to deterministic portion,Time,Performance,Deterministic performance change,Randomly selected time step,R

31、andom performance change,Deterministic Speed Increase,Speed increase =1.042265 + 0.021117 x Time_Interval,10,100,Speed mph,1000,10,000,1900,1920,1940,1960,1980,Wood & fabric,All metal,Subsonic jets,Aircraft speed prediction based on 1939,Supersonic jets,- 1 sigma,+ 1 sigma,Prediction line,Lower actu

32、al values in 1950s Politics - records were set by military A/C but actual speeds were not declared, only the increase above record,Probabilistic Lags,Several “similar” cases Time lag not constant Generate probability distribution for lag times Forecast is likelihood associated with various time lags

33、,Aluminium Aircraft Alloys Lag = avail time to first first flight of commercial A/C with that alloy. Q: if alloy introduced, how long to 1st flight?,Years Lag,Probability of Lag From Alloy Registration to First Flight,0,4,8,12,16,0,0.5,1.0,Individual Cumulative,Automotive Innovations,Innovation,Demo

34、 Car,Prestige Car,Mass Market,Years Lag,Probability of Lag: Demo Car to Mass MarketCar,0,5,10,15,20,0,0.5,1.0,Individual Cumulative,Years Lag,Probability of Lag: Prestige Car to Mass Market Car,0,5,10,15,20,0,0.5,1.0,Individual Cumulative,Multiple Leading Indicators,Use subsequent indicators to impr

35、ove estimate based on earlier indicators Later indicators constitute “new” information in Bayesian sense Apply Bayes Equation to improve estimate when new information obtained,Years to Prestige Car,Years to Mass Market,0,0,20,20,10,10,0.0,0.2,0.4,Probability Density,OHP,Joint Probability Density, Au

36、tomotive Innovations,Automotive Example (cont),Assume innovation appears in prestige car 8 years after demo carCompute new estimate of time lag from demo to mass market car,Stochastic Methods Summary,Methods available: stochastic projection probabilistic lags Methods not yet well developed Currently

37、 an area of research by technological forecasters,Measures of Technology,If you have multiple measuresScoring Models Technology Frontiers Expected Time Models,Scoring models,Algebraic combination of variables Usually in ratio form desirable variables in numerator (benefits) undesirable variables in

38、denominator (costs) Overriding variables multiply entire expression Variables weighted to reflect importance,Scoring Model for Breakfast Food,Vitamins Minerals Protein Calories Fibre,Taste moNetary cost pReparation time cLeanup time desirability of priZe in box,S = T (bV + cM + dP + eF)(1 + fZ)N R (

39、1 + gL)C,a,h,i,j,Steps in Forming Scoring Model,Identify factors to be included Weight the factors Construct the model,Identify Factors to be Included,Overriding variables if zero, all are zero Variables which can be traded all add up to 1.0 Optional variables ( 1 + xV ) if present, they influence i

40、f absent, no influence,Weight the Factors,Multiply tradable variables by weights Within a tradable group, sum of weights must be 1.0 Raise overriding variables to appropriate power Multiply optional variables by appropriate coefficient,Construct the model,Overriding variables multiply the expression

41、 Tradable variables grouped together Optional variables treated as one-variable groups Desirable variables in numerator Undesirable variables in denominator,Fighter Aircraft Performance Score,Score = Maneuverability x Availability x Range & Payload x Speed x Avionics x Weapons1 + Takeoff_Roll,Maneuv

42、erability = 0.3 x Instant_turn_rate + 0.3 x Sustained_turn_rate +0.4 x Climb_rateAvailability = 0.5 x Mean_time-between_failures + 0.5 x Flight_hours_per_maintenance_hourSpeed = 0.5 x Max_speed + 0.5 x Cruise_speedAvionics = 0.5 x Radar_range + 0.5 x Number_of_simultaneous_targetsWeapons = 0.2 x (#d

43、ogfight_missiles + #BVR_missiles + Range_of_Dogfight_missiles+ Range_of_BVR_Missiles + Guns)(0 = no, 1 = yes),2,2,1 E+2,1 E+3,1 E+4,1 E+5,1 E+6,1 E+7,Fighter Aircraft Performance Score,1940,1945,1955,1950,1960,1980,1985,1965,1970,1975,Score Log Plot,F15,F14,F18,F20,F80,F16,F5E,F4E,F5A,F8,F84,F86,F69

44、,F104,F102,F94,F101 F100,Tomcat,Eagle,Sabre,Starfighter,Technology Frontiers,see example,Effect of changing technology,T1,T2,T3,Whispertech,Expected Time Models,Identify relevant variables Regress time of introduction on variables,Expected Time for Jet Fighters,Year = 1938.740 + 5.431 x Maximum Mach

45、 Number + 4.036 x Mean Flying Hours Between Failures + 0.002 x Payload + 0.179 x Range of BVR Missiles,Other variables are not statistically relevant,1945,1955,1950,1960,1980,1985,1965,1970,1975,Predicted Year,F15,F14,F18,F20,F80,F16,F5E,F4E,F5A,F5,F84,F86,F69,F104,F106,F94,F101,Tomcat,Eagle,Sabre,S

46、tarfighter,Actual Year,1940,1945,1950,1955,1960,1965,1970,1975,1980,1985,F100,45 deg equality line,Measures Summary,Quantitative measures of technology needed Single-parameter measures must meet several criteria Multiple-parameter measures can be computed, depending on the nature of the technology,Product Development,What will competitors offer? Risk of technical failure Risk of early obsolescence Reverse regression: time on performance,